{"id":48832163,"url":"https://github.com/rajitsaha/agentbreeder","last_synced_at":"2026-04-14T21:01:00.605Z","repository":{"id":350984396,"uuid":"1177286901","full_name":"rajitsaha/agentbreeder","owner":"rajitsaha","description":"Stop wrangling AI agents. Start shipping them. One YAML file → any framework (LangGraph, OpenAI Agents, Claude SDK, CrewAI, Google ADK) → any cloud (AWS, GCP, Local) → governance automatic. RBAC, cost tracking, audit trail, org registry. Open source.","archived":false,"fork":false,"pushed_at":"2026-04-13T04:14:34.000Z","size":3497,"stargazers_count":0,"open_issues_count":18,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-04-13T05:22:10.082Z","etag":null,"topics":["agent-deployment","agent-framework","agent-orchestration","agentops","ai-agents","claude-sdk","crewai","developer-tools","fastapi","google-adk","governance","langgraph","llmops","mcp","mlops","multi-cloud","open-source","openai-agents","python","rbac"],"latest_commit_sha":null,"homepage":"https://rajitsaha.github.io/agentbreeder/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/rajitsaha.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":".github/CODEOWNERS","security":"SECURITY.md","support":null,"governance":null,"roadmap":"ROADMAP.md","authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-03-09T22:08:58.000Z","updated_at":"2026-04-13T04:14:39.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/rajitsaha/agentbreeder","commit_stats":null,"previous_names":["rajitsaha/agentbreeder"],"tags_count":1,"template":true,"template_full_name":null,"purl":"pkg:github/rajitsaha/agentbreeder","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rajitsaha%2Fagentbreeder","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rajitsaha%2Fagentbreeder/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rajitsaha%2Fagentbreeder/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rajitsaha%2Fagentbreeder/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/rajitsaha","download_url":"https://codeload.github.com/rajitsaha/agentbreeder/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rajitsaha%2Fagentbreeder/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31815080,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-14T18:05:02.291Z","status":"ssl_error","status_checked_at":"2026-04-14T18:05:01.765Z","response_time":153,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent-deployment","agent-framework","agent-orchestration","agentops","ai-agents","claude-sdk","crewai","developer-tools","fastapi","google-adk","governance","langgraph","llmops","mcp","mlops","multi-cloud","open-source","openai-agents","python","rbac"],"created_at":"2026-04-14T21:00:58.437Z","updated_at":"2026-04-14T21:01:00.593Z","avatar_url":"https://github.com/rajitsaha.png","language":"Python","readme":"\u003cdiv align=\"center\"\u003e\n\n# AgentBreeder\n\n### Stop wrangling agents. Start shipping them.\n\n**One YAML file. Any framework. Any cloud. Governance built in.**\n\n[![PyPI](https://img.shields.io/pypi/v/agentbreeder?color=blue\u0026label=PyPI)](https://pypi.org/project/agentbreeder/)\n[![PyPI Downloads](https://img.shields.io/pypi/dm/agentbreeder?color=green\u0026label=Downloads)](https://pypi.org/project/agentbreeder/)\n[![npm](https://img.shields.io/npm/v/@agentbreeder/sdk?color=red\u0026label=npm)](https://www.npmjs.com/package/@agentbreeder/sdk)\n[![Python](https://img.shields.io/pypi/pyversions/agentbreeder?color=blue)](https://pypi.org/project/agentbreeder/)\n[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE)\n[![CI](https://github.com/rajitsaha/agentbreeder/actions/workflows/ci.yml/badge.svg)](https://github.com/rajitsaha/agentbreeder/actions/workflows/ci.yml)\n[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg)](CONTRIBUTING.md)\n\n\u003cbr/\u003e\n\n[![LangGraph](https://img.shields.io/badge/LangGraph-supported-purple)](https://github.com/langchain-ai/langgraph)\n[![OpenAI Agents](https://img.shields.io/badge/OpenAI_Agents-supported-teal)](https://github.com/openai/openai-agents-python)\n[![Claude SDK](https://img.shields.io/badge/Claude_SDK-supported-orange)](https://docs.anthropic.com/en/docs/agents-and-tools/claude-code/sdk)\n[![CrewAI](https://img.shields.io/badge/CrewAI-supported-red)](https://github.com/crewAIInc/crewAI)\n[![Google ADK](https://img.shields.io/badge/Google_ADK-supported-4285F4)](https://github.com/google/adk-python)\n[![MCP](https://img.shields.io/badge/MCP-native-green)](https://modelcontextprotocol.io/)\n\n\u003cbr/\u003e\n\n[Quick Start](#quick-start) · [How It Works](#how-it-works) · [Install](#install) · [Features](#whats-implemented) · [CLI Reference](#cli) · [Docs](#documentation) · [Contributing](#contributing)\n\n\u003c/div\u003e\n\n---\n\nYour company has 47 AI agents. Nobody knows what they cost, who approved them, or which ones are still running. Three teams built the same summarizer. The security team hasn't audited any of them.\n\n**AgentBreeder fixes this.**\n\nWrite one `agent.yaml`. Run `agentbreeder deploy`. Your agent is live — with RBAC, cost tracking, audit trail, and org-wide discoverability. Automatic. Not optional.\n\n```\n╔═══════════════════════════════════════════════════════════════╗\n║                   AGENTBREEDER DEPLOY                         ║\n╠═══════════════════════════════════════════════════════════════╣\n║                                                               ║\n║  ✅  YAML parsed \u0026 validated                                  ║\n║  ✅  RBAC check passed (team: engineering)                    ║\n║  ✅  Dependencies resolved (3 tools, 1 prompt)                ║\n║  ✅  Container built (langgraph runtime)                      ║\n║  ✅  Deployed to GCP Cloud Run                                ║\n║  ✅  Health check passed                                      ║\n║  ✅  Registered in org registry                               ║\n║  ✅  Cost attribution: engineering / $0.12/hr                 ║\n║                                                               ║\n║  ENDPOINT: https://support-agent-a1b2c3.run.app              ║\n║  STATUS:   ✅ LIVE                                            ║\n║                                                               ║\n╚═══════════════════════════════════════════════════════════════╝\n```\n\n---\n\n## The Problem\n\nAI coding tools make it easy to **build** agents. Nobody has made it easy to **ship** them responsibly.\n\n| What happens today | What happens with AgentBreeder |\n|---|---|\n| Every framework has its own deploy story | One YAML, any framework, any cloud |\n| No RBAC — anyone deploys anything | RBAC validated before the first container builds |\n| No cost tracking — $40k surprise cloud bills | Cost attributed per team, per agent, per model |\n| No audit trail — \"who deployed that?\" | Every deploy logged with who, what, when, where |\n| No discoverability — duplicate agents everywhere | Org-wide registry — search before you build |\n| Governance is bolted on after the fact | Governance is a **structural side effect** of deploying |\n\n**Governance is not configuration. It is a side effect of the pipeline. There is no way to skip it.**\n\n---\n\n## How It Works\n\n```yaml\n# agent.yaml — this is the entire config\nname: customer-support-agent\nversion: 1.0.0\nteam: customer-success\nowner: alice@company.com\n\nframework: langgraph          # or: openai_agents, claude_sdk, crewai, google_adk, custom\n\nmodel:\n  primary: claude-sonnet-4\n  fallback: gpt-4o\n\ntools:\n  - ref: tools/zendesk-mcp    # pull from org registry\n  - ref: tools/order-lookup\n\ndeploy:\n  cloud: gcp                  # or: aws, local, kubernetes\n  scaling:\n    min: 1\n    max: 10\n```\n\n```bash\npip install agentbreeder\nagentbreeder deploy ./agent.yaml\n```\n\nThat's it. Eight atomic steps — parse, RBAC, resolve deps, build container, provision infra, deploy, health check, register. If any step fails, the entire deploy rolls back.\n\n---\n\n## Three Ways to Build\n\nAll three tiers compile to the same internal format. Same deploy pipeline. Same governance. No lock-in.\n\n| Tier | Who | How | Eject to |\n|------|-----|-----|----------|\n| **No Code** | PMs, analysts, citizen builders | Visual drag-and-drop canvas — pick model, tools, prompts from the registry | Low Code (view YAML) |\n| **Low Code** | ML engineers, DevOps | Write `agent.yaml` in any IDE | Full Code (`agentbreeder eject`) |\n| **Full Code** | Senior engineers, researchers | Python/TS SDK with full programmatic control | — |\n\n```python\n# Full Code SDK — builder pattern\nfrom agenthub import Agent\n\nagent = (\n    Agent(\"support-agent\", version=\"1.0.0\", team=\"eng\")\n    .with_model(primary=\"claude-sonnet-4\", fallback=\"gpt-4o\")\n    .with_tools([\"tools/zendesk-mcp\", \"tools/order-lookup\"])\n    .with_prompt(system=\"You are a helpful customer support agent.\")\n    .with_deploy(cloud=\"gcp\", min_scale=1, max_scale=10)\n)\nagent.deploy()\n```\n\n---\n\n## What's Implemented\n\n### Frameworks\n\n| Framework | Status | Runtime |\n|-----------|--------|---------|\n| LangGraph | ✅ | `engine/runtimes/langgraph.py` |\n| OpenAI Agents SDK | ✅ | `engine/runtimes/openai_agents.py` |\n| Claude SDK (Anthropic) | ✅ | `engine/runtimes/claude_sdk.py` |\n| CrewAI | ✅ | `engine/runtimes/crewai.py` |\n| Google ADK | ✅ | `engine/runtimes/google_adk.py` |\n| Custom (bring your own) | ✅ | `engine/runtimes/custom.py` |\n\n### Cloud Targets\n\n| Target | Status | Deployer |\n|--------|--------|----------|\n| Local (Docker Compose) | ✅ | `engine/deployers/docker_compose.py` |\n| GCP Cloud Run | ✅ | `engine/deployers/gcp_cloudrun.py` |\n| AWS ECS Fargate | 🔲 Planned | |\n| Kubernetes | 🔲 Planned | |\n\n### LLM Providers (6 providers + fallback chains)\n\n| Provider | Status |\n|----------|--------|\n| Anthropic (Claude) | ✅ |\n| OpenAI (GPT-4o, o1, etc.) | ✅ |\n| Google (Gemini) | ✅ |\n| Ollama (local models) | ✅ |\n| LiteLLM gateway (100+ models) | ✅ |\n| OpenRouter | ✅ |\n\n### Secrets Backends\n\n| Backend | Status |\n|---------|--------|\n| Environment variables / `.env` | ✅ |\n| AWS Secrets Manager | ✅ |\n| GCP Secret Manager | ✅ |\n| HashiCorp Vault | ✅ |\n\n### Platform Features (30+ shipped)\n\n| Feature | Status |\n|---------|--------|\n| Org-wide agent registry | ✅ |\n| Visual agent builder (ReactFlow canvas) | ✅ |\n| Multi-agent orchestration (6 strategies) | ✅ |\n| Visual orchestration canvas | ✅ |\n| A2A (Agent-to-Agent) protocol | ✅ |\n| MCP server hub + sidecar injection | ✅ |\n| Agent evaluation framework | ✅ |\n| Cost tracking (per team / agent / model) | ✅ |\n| RBAC + team management | ✅ |\n| Full audit trail | ✅ |\n| Distributed tracing (OpenTelemetry) | ✅ |\n| AgentOps fleet dashboard | ✅ |\n| Community marketplace + templates | ✅ |\n| Git workflow (PR create → review → publish) | ✅ |\n| Prompt builder + test panel | ✅ |\n| RAG index builder | ✅ |\n| Memory configuration | ✅ |\n| Tool sandbox execution | ✅ |\n| Interactive chat playground | ✅ |\n| API versioning (v1 stable, v2 preview) | ✅ |\n| Python SDK | ✅ |\n| TypeScript SDK | ✅ |\n| Tier mobility (`agentbreeder eject`) | ✅ |\n\n---\n\n## Orchestration\n\nSix strategies. Define in YAML or the visual canvas — both compile to the same pipeline.\n\n```yaml\n# orchestration.yaml\nname: support-pipeline\nversion: \"1.0.0\"\nteam: customer-success\nstrategy: router       # router | sequential | parallel | hierarchical | supervisor | fan_out_fan_in\n\nagents:\n  triage:\n    ref: agents/triage-agent\n    routes:\n      - condition: billing\n        target: billing\n      - condition: default\n        target: general\n  billing:\n    ref: agents/billing-agent\n    fallback: general\n  general:\n    ref: agents/general-agent\n\nshared_state:\n  type: session_context\n  backend: redis\n\ndeploy:\n  target: gcp\n```\n\nOr programmatically with the SDK:\n\n```python\nfrom agenthub import Orchestration\n\npipeline = (\n    Orchestration(\"support-pipeline\", strategy=\"router\", team=\"eng\")\n    .add_agent(\"triage\",  ref=\"agents/triage-agent\")\n    .add_agent(\"billing\", ref=\"agents/billing-agent\")\n    .add_agent(\"general\", ref=\"agents/general-agent\")\n    .with_route(\"triage\", condition=\"billing\", target=\"billing\")\n    .with_route(\"triage\", condition=\"default\", target=\"general\")\n    .with_shared_state(state_type=\"session_context\", backend=\"redis\")\n)\npipeline.deploy()\n```\n\n---\n\n## Install\n\n### PyPI (recommended)\n\n```bash\n# Full CLI + API server + engine\npip install agentbreeder\n\n# Lightweight Python SDK only (for programmatic agent definitions)\npip install agentbreeder-sdk\n```\n\n### npm (TypeScript / JavaScript)\n\n```bash\nnpm install @agentbreeder/sdk\n```\n\n```typescript\nimport { Agent } from \"@agentbreeder/sdk\";\n\nconst agent = new Agent(\"customer-support\", { version: \"1.0.0\", team: \"eng\" })\n  .withModel({ primary: \"claude-sonnet-4\", fallback: \"gpt-4o\" })\n  .withTool({ ref: \"tools/zendesk-mcp\" })\n  .withDeploy({ cloud: \"aws\", region: \"us-east-1\" });\n\nawait agent.deploy();\n```\n\n### Homebrew (macOS / Linux)\n\n```bash\nbrew tap rajitsaha/agentbreeder\nbrew install agentbreeder\n```\n\n### Docker\n\n```bash\n# API server\ndocker pull rajits/agentbreeder-api\ndocker run -p 8000:8000 rajits/agentbreeder-api\n\n# Dashboard\ndocker pull rajits/agentbreeder-dashboard\ndocker run -p 3001:3001 rajits/agentbreeder-dashboard\n\n# CLI (for CI/CD pipelines)\ndocker pull rajits/agentbreeder-cli\ndocker run rajits/agentbreeder-cli deploy agent.yaml --target gcp\n```\n\n---\n\n## Quick Start\n\n```bash\npip install agentbreeder\n\n# Scaffold your first agent (interactive wizard — pick framework, cloud, model)\nagentbreeder init\n\n# Validate the config\nagentbreeder validate agent.yaml\n\n# Deploy locally\nagentbreeder deploy agent.yaml --target local\n```\n\n**Or run the full platform locally:**\n\n```bash\ngit clone https://github.com/rajitsaha/agentbreeder.git\ncd agentbreeder\n\npython -m venv venv \u0026\u0026 source venv/bin/activate\npip install -e \".[dev]\"\ncp .env.example .env\n\n# Start postgres + redis + API + dashboard\ndocker compose -f deploy/docker-compose.yml up -d\n```\n\nDashboard: `http://localhost:3001` · API: `http://localhost:8000` · API Docs: `http://localhost:8000/docs`\n\nSee [docs/quickstart.md](docs/quickstart.md) for the full guide.\n\n---\n\n## CLI\n\n24 commands. Everything you need from scaffold to teardown.\n\n```bash\nagentbreeder init              # Scaffold a new agent project (interactive wizard)\nagentbreeder validate          # Validate agent.yaml without deploying\nagentbreeder deploy            # Deploy an agent (the core command)\nagentbreeder up / down         # Start / stop the full local platform stack\nagentbreeder status            # Show deploy status\nagentbreeder logs \u003cname\u003e       # Tail agent logs\nagentbreeder list              # List agents / tools / models / prompts\nagentbreeder describe \u003cname\u003e   # Show detail for a registry entity\nagentbreeder search \u003cquery\u003e    # Search across the entire registry\nagentbreeder chat \u003cname\u003e       # Interactive chat with a deployed agent\nagentbreeder eval              # Run evaluations against golden datasets\nagentbreeder eject             # Eject from YAML to Full Code SDK\nagentbreeder submit            # Create a PR for review\nagentbreeder review            # Review / approve / reject a submission\nagentbreeder publish           # Merge approved PR and publish to registry\nagentbreeder provider          # Manage LLM provider connections\nagentbreeder secret            # Manage secrets across backends (env, AWS, GCP, Vault)\nagentbreeder scan              # Discover MCP servers and LiteLLM models\nagentbreeder template          # Manage agent templates\nagentbreeder orchestration     # Multi-agent orchestration commands\nagentbreeder teardown          # Remove a deployed agent and clean up resources\n```\n\nSee [docs/cli-reference.md](docs/cli-reference.md) for full usage and flags.\n\n---\n\n## Architecture\n\n```\nDeveloper                    AgentBreeder Platform                  Cloud\n                                     ┌──────────────┐\nagent.yaml  ──▶  [ CLI ]  ──▶  │  API Server  │  ──▶  [ Engine ]  ──▶  AWS / GCP / Local\n                                     └──────┬───────┘         │\n                                            │                 ▼\n                                     ┌──────▼───────┐  ┌─────────────────┐\n                                     │  PostgreSQL   │  │ Container Build │\n                                     │  (Registry)   │  │  + MCP Sidecar  │\n                                     └──────┬───────┘  └─────────────────┘\n                                            │\n                                     ┌──────▼───────┐\n                                     │    Redis      │\n                                     │   (Queue)     │\n                                     └──────────────┘\n```\n\n**Deploy pipeline** — 8 atomic steps. If any fails, the entire deploy rolls back:\n\n1. Parse \u0026 validate YAML\n2. RBAC check (fail fast if unauthorized)\n3. Dependency resolution (tools, prompts, models from registry)\n4. Container build (framework-specific Dockerfile)\n5. Infrastructure provision (Pulumi)\n6. Deploy \u0026 health check\n7. Auto-register in org registry\n8. Return endpoint URL\n\n---\n\n## Project Structure\n\n```\nagentbreeder/\n├── api/                # FastAPI backend — 25 route modules, services, models\n├── cli/                # CLI — 24 commands (Typer + Rich)\n├── engine/\n│   ├── config_parser.py       # YAML parsing + JSON Schema validation\n│   ├── builder.py             # 8-step atomic deploy pipeline\n│   ├── orchestrator.py        # Multi-agent orchestration engine\n│   ├── providers/             # LLM providers (Anthropic, OpenAI, Google, Ollama)\n│   ├── runtimes/              # Framework builders (LangGraph, OpenAI Agents)\n│   ├── deployers/             # Cloud deployers (Docker Compose, GCP Cloud Run)\n│   ├── secrets/               # Secrets backends (env, AWS, GCP, Vault)\n│   ├── a2a/                   # Agent-to-Agent protocol\n│   └── mcp/                   # MCP server packaging\n├── registry/           # Catalog services — agents, tools, models, prompts, templates\n├── sdk/python/         # Python SDK (pip install agentbreeder-sdk)\n├── sdk/typescript/     # TypeScript SDK (npm install @agentbreeder/sdk)\n├── connectors/         # LiteLLM, OpenRouter, MCP scanner\n├── dashboard/          # React + TypeScript + Tailwind\n├── tests/              # 2,378 unit tests + Playwright E2E\n└── examples/           # Working examples per framework + orchestration\n```\n\n---\n\n## Documentation\n\n| Doc | Description |\n|-----|-------------|\n| [Quickstart](docs/quickstart.md) | Local setup in under 10 minutes |\n| [CLI Reference](docs/cli-reference.md) | All 24 commands with flags and examples |\n| [agent.yaml Reference](docs/agent-yaml.md) | Full configuration field reference |\n| [orchestration.yaml Reference](docs/orchestration-yaml.md) | Multi-agent pipeline config |\n| [Orchestration SDK](docs/orchestration-sdk.md) | Python/TS SDK for complex workflows |\n| [API Stability](docs/api-stability.md) | Versioning and deprecation policy |\n| [Local Development](docs/local-development.md) | Contributor setup guide |\n| [ARCHITECTURE.md](ARCHITECTURE.md) | Technical deep-dive |\n| [ROADMAP.md](ROADMAP.md) | Release plan and milestone status |\n| [CONTRIBUTING.md](CONTRIBUTING.md) | How to contribute |\n\n---\n\n## \"How is this different from LangGraph / CrewAI / Mastra / OpenAI Agents?\"\n\n**It's not the same category.**\n\nAgent SDKs help you **build** an agent. AgentBreeder helps you **ship, govern, and operate** it. You use them together — not instead of each other.\n\n```\n┌─────────────────────────────────────────────────────────────────────┐\n│                        What you do today                            │\n│                                                                     │\n│   LangGraph / CrewAI / Mastra / OpenAI Agents / Claude SDK          │\n│   ──────────────────────────────────────────────────────────         │\n│   Build the agent            ✅  Great at this                      │\n│   Deploy it somewhere        🤷  You figure it out                  │\n│   Track who deployed what    🤷  You figure it out                  │\n│   Control who can deploy     🤷  You figure it out                  │\n│   Track costs per team       🤷  You figure it out                  │\n│   Audit every deploy         🤷  You figure it out                  │\n│   Discover agents across org 🤷  You figure it out                  │\n│   Multi-cloud portability    🤷  You figure it out                  │\n└─────────────────────────────────────────────────────────────────────┘\n\n┌─────────────────────────────────────────────────────────────────────┐\n│                     What changes with AgentBreeder                  │\n│                                                                     │\n│   LangGraph / CrewAI / Mastra / OpenAI Agents / Claude SDK          │\n│                          +                                          │\n│                    AgentBreeder                                     │\n│   ──────────────────────────────────────────────────────────         │\n│   Build the agent            ✅  Your SDK does this (unchanged)     │\n│   Deploy it somewhere        ✅  `agentbreeder deploy` — any cloud  │\n│   Track who deployed what    ✅  Automatic audit trail              │\n│   Control who can deploy     ✅  RBAC validated before build starts │\n│   Track costs per team       ✅  Per team / agent / model           │\n│   Audit every deploy         ✅  Every deploy logged                │\n│   Discover agents across org ✅  Org-wide searchable registry      │\n│   Multi-cloud portability    ✅  Same YAML → AWS, GCP, or local    │\n└─────────────────────────────────────────────────────────────────────┘\n```\n\n**Think of it this way:**\n\n| Analogy | Build tool | Ship/Operate tool |\n|---------|-----------|-------------------|\n| Code | You write Python | Docker + Kubernetes deploys it |\n| Infrastructure | Terraform defines it | CI/CD ships it |\n| **Agents** | **LangGraph / CrewAI / Mastra builds it** | **AgentBreeder ships and governs it** |\n\nAgentBreeder is to agents what **Docker + Kubernetes is to microservices** — the deployment, orchestration, and operations layer. Your agent framework is the application; AgentBreeder is the platform.\n\n### The real comparison\n\n| | Agent SDKs | Agent Platforms | AgentBreeder |\n|---|---|---|---|\n| | LangGraph, CrewAI, Mastra, OpenAI Agents, Claude SDK, Google ADK | Vertex AI Agent Builder, Amazon Bedrock Agents, Azure AI Agent Service | |\n| **Purpose** | Build agents | Build + deploy (their way) | Ship + govern + operate (any agent, any cloud) |\n| **Framework lock-in** | ✅ You're locked in | ✅ You're locked in | ❌ Bring any framework |\n| **Cloud lock-in** | N/A | ✅ Their cloud only | ❌ AWS, GCP, local, K8s |\n| **Works with your existing agents** | N/A | ❌ Rewrite required | ✅ Wrap in `agent.yaml`, deploy |\n| **Governance** | ❌ Not their problem | ⚠️ Partial, platform-specific | ✅ RBAC, audit, cost — automatic |\n| **Org-wide registry** | ❌ | ❌ | ✅ Every deploy auto-registered |\n| **Cost attribution** | ❌ | ⚠️ Project-level | ✅ Per team / agent / model |\n| **Multi-agent orchestration** | ⚠️ Framework-specific | ⚠️ Limited | ✅ 6 strategies, cross-framework |\n| **MCP native** | ⚠️ Varies | ❌ | ✅ Hub + sidecar injection |\n| **Open source** | ✅ Most are | ❌ All proprietary | ✅ Apache 2.0 |\n| **Vendor lock-in** | Framework | Cloud vendor | **None** |\n\n### vs. Managed Agent Platforms (detailed)\n\nEvery cloud vendor now has an agent platform. They all share the same tradeoff: **convenience in exchange for lock-in.** AgentBreeder gives you the convenience without the lock-in.\n\n| | AWS Bedrock Agents | Vertex AI Agent Builder | Azure AI Agent Service | Databricks AgentBricks | Claude Managed Agents | AgentBreeder |\n|---|---|---|---|---|---|---|\n| **Cloud** | AWS only | GCP only | Azure only | Databricks only | Anthropic only | Any cloud + local |\n| **Frameworks** | Bedrock SDK | Vertex SDK / ADK | Azure SDK | Mosaic Agent Framework | Claude SDK | LangGraph, CrewAI, OpenAI, Claude, ADK, custom |\n| **Bring your own agent code** | ❌ Rewrite in their SDK | ❌ Rewrite in their SDK | ❌ Rewrite in their SDK | ❌ Rewrite in their SDK | ❌ Claude only | ✅ Any framework, wrap in YAML |\n| **Multi-cloud deploy** | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ Same config → any target |\n| **Migrate away** | Painful (SDK lock-in) | Painful (SDK lock-in) | Painful (SDK lock-in) | Painful (platform lock-in) | Painful (API lock-in) | Trivial (it's your code + YAML) |\n| **RBAC** | IAM (AWS-specific) | IAM (GCP-specific) | RBAC (Azure-specific) | Unity Catalog | API keys | Framework-agnostic RBAC |\n| **Cost tracking** | CloudWatch + Cost Explorer | GCP Billing | Azure Cost Management | DBU tracking | Usage dashboard | Built-in: per team / agent / model |\n| **Org-wide registry** | ❌ | ❌ | ❌ | Unity Catalog (models) | ❌ | ✅ Agents, tools, prompts, models |\n| **Multi-agent orchestration** | Step Functions (manual) | Limited | Limited | Limited | Tool use (single agent) | 6 strategies (router, sequential, parallel, hierarchical, supervisor, fan-out) |\n| **MCP support** | ❌ | ❌ | ❌ | ❌ | ✅ Native | ✅ Hub + sidecar injection |\n| **Open source** | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ Apache 2.0 |\n| **Data residency** | Their cloud | Their cloud | Their cloud | Their platform | Their API | **Your infrastructure** |\n| **Pricing** | Pay per invocation + hosting | Pay per invocation + hosting | Pay per invocation + hosting | DBU consumption | Pay per token | Free (you pay your own cloud costs) |\n\n**The pattern:** Every managed platform wants you to rewrite your agent in their SDK, deploy to their cloud, and pay their markup. When you want to switch, you rewrite everything.\n\n**AgentBreeder's position:** Your agent code stays yours. Your cloud stays yours. Your data stays yours. AgentBreeder is the deploy + governance layer that works across all of them.\n\n### What AgentBreeder is NOT\n\n- **Not a replacement for LangGraph.** Use LangGraph to build your agent's logic. Use AgentBreeder to deploy, govern, and operate it.\n- **Not a replacement for CrewAI.** Build your crew in CrewAI. Ship it with AgentBreeder.\n- **Not a replacement for Mastra.** Build your TypeScript agent in Mastra. Wrap it in `agent.yaml`. Deploy it anywhere.\n- **Not competing with Bedrock / Vertex / Azure.** They're managed services. AgentBreeder is open-source infrastructure. Different model, different tradeoffs.\n- **Not another agent framework.** We don't have opinions about how you build agents. We have opinions about how you ship them responsibly.\n- **Not a managed service.** It's open-source infrastructure you run. No vendor lock-in. No data leaves your cloud.\n\n---\n\n## Contributing\n\nHigh-impact areas where contributions are especially welcome:\n\n- **AWS ECS deployer** — `engine/deployers/aws_ecs.py` — most requested cloud target\n- **Framework runtimes** — additional frameworks beyond the six currently supported\n- **Agent templates** — starter templates for common use cases\n- **Connectors** — Datadog, Grafana, and other observability integrations\n\nSee [CONTRIBUTING.md](CONTRIBUTING.md) for setup instructions and guidelines.\n\n---\n\n## Community\n\n- [GitHub Issues](https://github.com/rajitsaha/agentbreeder/issues) — bugs and feature requests\n- [GitHub Discussions](https://github.com/rajitsaha/agentbreeder/discussions) — questions and show \u0026 tell\n\n---\n\n## License\n\nApache License 2.0 — see [LICENSE](LICENSE).\n\n---\n\n\u003cdiv align=\"center\"\u003e\n\n### Built by [Rajit Saha](https://www.linkedin.com/in/rajsaha/)\n\nTech executive · 20+ years building enterprise data \u0026 ML platforms · Udemy, LendingClub, VMware, Yahoo\n\n[![LinkedIn](https://img.shields.io/badge/LinkedIn-Rajit_Saha-blue?logo=linkedin)](https://www.linkedin.com/in/rajsaha/)\n[![GitHub](https://img.shields.io/github/followers/rajitsaha?label=Follow\u0026style=social)](https://github.com/rajitsaha)\n\n\u003cbr/\u003e\n\n**If AgentBreeder saves you time, [star the repo](https://github.com/rajitsaha/agentbreeder) and share it with your team.**\n\n\u003c/div\u003e\n","funding_links":[],"categories":[],"sub_categories":[],"project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Frajitsaha%2Fagentbreeder","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Frajitsaha%2Fagentbreeder","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Frajitsaha%2Fagentbreeder/lists"}